On Wed, Sep 7, 2011 at 8:14 AM, Jörg Schulz <jorg.schulz@lnu.se> wrote:

Try to implement shinken in our enviroment , we have physical 2 sites , on each site we have our own nagios server , the servers checks the their  own site.
Now we have to connect with each other. Both should have redundance/loadbalancing/failover.
Which is the easiet way , because we need the merlin (op5) db for different features.

-------Site1------- <------->  ------Site2------
        |                           |
    Merlin                merlin
  Local Network check   Local Network check


How could the config looks like for this scenario

I think the best way for this is to define two realms with both :
* 2 schedulers, 2 pollers, 2 reactionners, 2 brokers (where you will put the merlin module) and 2 receivers (for passive checks, but it can be the main arbiter that manage it), one arbiter in each, one site is the master arbiter, the other is the slave one.

You will be able to scale one site without problem if you need, and you still got one common configuration place. It's fully HA, and even if you lsot the inter-link, the slave arbiter will just resend the same configuration, and when the master will came back it will just take the lead. With it you just need to tag your hosts or hostgroups with the realm name.

You can also take a cheaper installation with only one scheduler in each site and do the "site" management with poller tags, but it will ask more network trafic between your sites. If it's not a problem, it can be an easier solution :)




Using storage to extend the benefits of virtualization and iSCSI
Virtualization increases hardware utilization and delivers a new level of
agility. Learn what those decisions are and how to modernize your storage
and backup environments for virtualization.
Shinken-devel mailing list